feat: add batch replan mode and comprehensive skill system

- Add batch processing mode to /task:replan command
  - Support for verification report input
  - TodoWrite integration for progress tracking
  - Automatic backup management
- Enhance /workflow:action-plan-verify with batch remediation
  - Save verification report to .process directory
  - Provide batch replan command suggestions
- Add comprehensive skill documentation
  - Codex: autonomous development workflows
  - Gemini/Qwen: code analysis and documentation
  - Context-search: strategic context gathering
  - Prompt-enhancer: ambiguous prompt refinement
- Clean up CLAUDE.md strategy references

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>
This commit is contained in:
catlog22
2025-10-17 11:44:37 +08:00
parent 80196cc0a0
commit b366924ae6
24 changed files with 4231 additions and 51 deletions

View File

@@ -0,0 +1,100 @@
---
name: Prompt Enhancer
description: Systematically enhance ambiguous user prompts by combining session memory with codebase analysis. Use when user input is unclear, vague, or lacks technical specificity (e.g., "fix", "improve", "clean up") or affects multiple modules or critical systems.
allowed-tools: Bash, Read, Glob, Grep
---
# Prompt Enhancer
## Overview
Transforms ambiguous user requests into actionable technical specifications through semantic analysis and session memory integration.
**Core Capability**: Vague intent → Structured specification
## Enhancement Process
### Step 1: Semantic Analysis
Analyze user input to identify:
- **Intent keywords**: fix, improve, add, refactor, update, migrate
- **Technical scope**: single file vs multi-module
- **Domain context**: auth, payment, security, API, UI, database
- **Implied requirements**: performance, security, testing, documentation
### Step 2: Memory Analysis
Extract from conversation history:
- **Technical context**: Previous discussions, decisions, implementations
- **Known patterns**: Identified code patterns, architecture decisions
- **Current state**: What's been built, what's in progress
- **Dependencies**: Related modules, integration points
- **Constraints**: Security requirements, backward compatibility
### Step 3: Context Integration
Combine semantic and memory analysis to determine:
- **Precise intent**: Specific technical goal
- **Required actions**: Implementation steps with file references
- **Critical constraints**: Security, compatibility, testing requirements
- **Missing information**: What needs clarification
## Output Structure
Every enhanced prompt must follow this format:
```
INTENT: [Clear technical goal]
CONTEXT: [Session memory + semantic analysis]
ACTION: [Numbered implementation steps]
ATTENTION: [Critical constraints]
```
**Field Descriptions**:
- **INTENT**: One-sentence technical goal derived from semantic analysis
- **CONTEXT**: Session memory findings + semantic domain analysis
- **ACTION**: Numbered steps with specific file/module references
- **ATTENTION**: Critical constraints, security, compliance, tests
## Semantic Patterns
### Intent Translation
| User Input | Semantic Intent | Focus |
|------------|----------------|-------|
| "fix" + vague target | Debug and resolve | Root cause → preserve behavior |
| "improve" + no metrics | Enhance/optimize | Performance/readability |
| "add" + feature name | Implement feature | Integration + edge cases |
| "refactor" + module | Restructure | Maintain behavior |
| "update" + version | Modernize | Version compatibility |
### Scope Detection
**Single-file scope**:
- "fix button", "add validation", "update component"
- Use session memory only
**Multi-module scope** (>3 modules):
- "add authentication", "refactor payment", "migrate database"
- Analyze dependencies and integration points
**System-wide scope**:
- "improve performance", "add logging", "update security"
- Consider cross-cutting concerns
## Key Principles
1. **Memory First**: Check session memory before assumptions
2. **Semantic Precision**: Extract exact technical intent from vague language
3. **Context Reuse**: Build on previous understanding
4. **Clear Output**: Always structured format
5. **Avoid Duplication**: Reference context, don't repeat
## Best Practices
- **Semantic analysis**: Identify domain, scope, and intent keywords
- **Memory integration**: Extract all relevant context from conversation
- **Structured output**: Always use INTENT/CONTEXT/ACTION/ATTENTION format
- **Actionable steps**: Specific files, clear execution order
- **Critical constraints**: Security, compatibility, testing requirements